It reads a developer's actual code to build a living map of what they're truly great at, instead of trusting what their resume says.

Skillsync deploys transformer models (including architectures like SMITH) and custom NLP pipelines to ingest and analyze public code repositories at scale. The system parses pull requests, code reviews, commit messages, and inline comments to extract granular technical skills, domain expertise, and collaboration behaviors. These signals are assembled into a multi-dimensional skill graph for each developer, capturing not just language proficiency but architectural decision-making, code quality patterns, and cross-project influence. This approach surfaces hidden experts—developers who may lack polished resumes or LinkedIn profiles but demonstrate elite capability through their actual work. The skill graph updates dynamically as new contributions are made, providing recruiters with a living, evidence-based talent map rather than a static snapshot.

It's like judging a chef by tasting every dish they've ever cooked instead of just reading their food blog.

It's like having a career counselor who's read every job posting, every codebase you've touched, and every industry trend—then tells you exactly what to learn next.

Skillsync's AskJill system combines large language models with knowledge graph reasoning engines (built on MeTTa and uAgents frameworks) to create a multi-agent mentorship experience. The system ingests a developer's skill graph, cross-references it against labor market data, emerging technology trends, and organizational needs, then generates personalized career development recommendations. Multiple AI agents collaborate—one analyzing skill gaps, another tracking market demand, and a third identifying optimal mentorship matches within or across organizations. Explainable AI techniques ensure that every recommendation comes with transparent reasoning, so users understand why a particular path or mentor is suggested. This moves beyond simple job matching into proactive, evidence-based career intelligence that adapts as both the individual and the market evolve.

It's like GPS for your career—except it also knows every road being built next year and which ones have the best scenery.

It finds the people who quietly make every project better by studying how they review and improve other people's code, not just their own.

Skillsync trains ML classification and NLP sentiment models on the full lifecycle of pull request activity—submissions, reviews, comment threads, approval patterns, and revision cycles. The system identifies behavioral signatures that distinguish true domain experts: developers who consistently improve code quality through reviews, catch subtle bugs, mentor junior contributors through comments, and shape architectural decisions across repositories. By analyzing these interaction patterns at scale, Skillsync surfaces experts who may never appear on leaderboards or conference stages but whose influence is deeply embedded in the codebases they touch. This is particularly valuable for recruiting in niche technical domains (e.g., cryptography, compiler design, embedded systems) where traditional sourcing methods fail to reach the most impactful contributors. The models are continuously retrained as new contribution data flows in, ensuring identification accuracy improves over time.

It's like finding the best basketball coach by watching game film of every player they've ever trained, instead of checking their win-loss record.